Case study / Document intelligence
From PDFs to verifiable answers.
Shivru AI connects medical-legal records with structured extraction, retrieval, and source-linked findings.
IngestPrivate PDFs
ExtractText + OCR
StructureTyped facts
Page-bound
Page-bound
GroundSource checks
Excerpt support
Excerpt support
AnswerCited findings
Human review
Human review
The problem
Medical-legal review involves information spread across documents, clinical events, bills, and supporting records. Useful AI assistance needs more than a fluent answer: a reviewer needs to see the underlying evidence.
My contribution
As a co-founder, I contribute to OCR, structured extraction, medical-legal retrieval, and evidence citations. My work supports medical chronology, billing reconciliation, missing-record checks, and citation verification.
A workflow built around evidence
- Ingest: Bring medical documents into the review workflow.
- Extract: Use OCR and text extraction to make document content available.
- Structure: Organize extracted facts while retaining their page context.
- Ground: Connect findings and retrieved content with evidence citations.
- Review: Support chronology, record checks, and answers that a person can verify.
What the system supports
- Medical chronology and event organization
- Billing reconciliation
- Missing-record checks
- Source-linked retrieval and citation verification
The diagram is a conceptual overview of the workflow. Human review remains part of interpreting the records and checking the evidence.